Causal Analysis and Latent Class Models: An Application to a Communication Hierarchy of Effects Model

Abstract

The authors illustrate the use of latent structure analysis to test, in a confirmatory sense, causal hypotheses in an experimental design setting. Two latent factors conceptualized as arousal and yielding are hypothesized to explain the linkages in a communication hierarchy of effects model. A stagewise analysis is proposed which can help in the analysis of multiway tables characterized by cell sparseness. In addition, the analysis addresses the case of polytomous latent factors and polytomous manifest variables.